Analytical and Bioanalytical Chemistry

, Volume 394, Issue 4, pp 1117–1124

MicroRNA detection by microarray

Authors

  • Wei Li
    • State Key Laboratory of Molecular Biology, Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological SciencesChinese Academy of Sciences
    • State Key Laboratory of Molecular Biology, Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological SciencesChinese Academy of Sciences
Review

DOI: 10.1007/s00216-008-2570-2

Cite this article as:
Li, W. & Ruan, K. Anal Bioanal Chem (2009) 394: 1117. doi:10.1007/s00216-008-2570-2

Abstract

MicroRNAs (miRNAs) are a class of small noncoding RNAs ∼22 nt in length that regulate gene expression and play fundamental roles in multiple biological processes, including cell differentiation, proliferation and apoptosis as well as disease processes. The study of miRNA has thus become a rapidly emerging field in life science. The detection of miRNA expression is a very important first step in miRNA exploration. Several methodologies, including cloning, northern blotting, real-time RT-PCR, microRNA arrays and ISH (in situ hybridization), have been developed and applied successfully in miRNA profiling. This review discusses the main existing microRNA detection technologies, while emphasizing microRNA arrays.

Keywords

MicroRNAMicroRNA detectionMicroRNA microarrayMicroRNA labeling

Introduction

MicroRNAs are a class of ∼22 nt long noncoding RNAs. They are derived from the cleavage of hairpin precursors by Dicer (a member of the RNAase III family), and are evolutionarily conserved [13]. The high conservation of microRNA sequences highlights the significance of their function. In animal cells, microRNAs inhibit the translation of target genes by binding with imperfect complementarity to multiple sites in the 3′ untranslated region (UTR) of target mRNAs [46]. A single microRNA can regulate the expression of many target genes, and a target gene can also be regulated by several microRNAs [7, 8]. Early studies suggested that microRNA genes account for about 1% of the genome in a species [9, 10]. However, according to the study early in 2005, the number of human microRNAs is greater than 1000 [11], and over one-third of human genes appear to be conserved microRNA targets [12].

At present, microRNAs are known to be involved in many biological functions, including the timing of early larval development transitions [13, 14], left/right asymmetry of chemoreceptor expression in nematodes [15], cell proliferation and apoptosis in insects [16], and hematopoietic differentiation in mammals [17], as well as leaf development [18], flowering transition timing [19], and flower development of Arabidopsis [20]. In addition, microRNAs are associated with many major diseases in humans. For example, miR15 and miR16 are associated with human B-cell chronic lymphatic leukemia (B-CLL) [21], reduced expression of let-7 in lung cancer leads to shortened postoperative survival [22], miR155 is associated with B-cell lymphomas [23], meanwhile miR-143 and miR-145 display reduced expression levels in colorectal tumor cells compared with normal colonic cells [24]. Moreover, miR-1 is overexpressed in individuals with coronary artery disease [25], and overexpression of MiR-17-92 in the lymphocytes of mice results in lymphoproliferative disease and autoimmunity [26]. All of these findings indicate that microRNA expression levels are closely associated with developmental stages and physiological states as well as disease processes. From these findings, it has also been established that microRNA expression detection and analysis is a basic and preliminary procedure in most miRNA studies. So far, several main techniques for detecting and analyzing miRNA expression, such as northern blotting, real-time RT-PCR, cloning and microarrays, have been developed and applied in miRNA exploration. Each technique has its advantages along with some limitations over its application at the moment. A suitable method should be chosen based on the requirements of the investigation and the experimental conditions in order to get accurate information on miRNA expression. This review will summarize and discuss current detection techniques for miRNA expression, with emphasis placed on miRNA microarrays, including probe design, sample preparation and miRNA labeling.

Common methodologies for studying microRNA expression

Current methods widely used in the study of microRNA expression mainly include microRNA cloning, northern blotting, real-time RT-PCR, microRNA arrays, and ISH. MicroRNA cloning is mainly used to discover new microRNAs, and the cloning frequency of microRNAs can, to some extent, reflect their relative abundance [27, 28]. Several different methods have been applied for microRNA cloning [29, 30], but their basic principles and procedures are similar. First, small RNA molecules are isolated by denaturing polyacrylamide gel electrophoresis, the 3′ and 5′ ends of the small RNA molecules are ligated with adaptor sequences that contain restriction sites, and PCR primers are then designed based on the adaptor sequences. Then, the PCR products obtained from RT-PCR amplification are transferred into the vectors for further cloning and sequencing analysis. Northern blotting, which can reflect the microRNA expression profile more accurately than cloning, is a routine method applied to the study of microRNA expression. Sempere et al. employed northern blotting to study the expression of 119 microRNA in the cerebrums, livers, muscles and lungs of mice [31]. The basic procedure of microRNA northern blotting involves: 1) fractionating small RNA molecules by denaturing polyacrylamide gel electrophoresis; 2) transferring RNA from the gel onto membrane; 3) fixing RNA onto membrane through a crosslinking procedure; 4) hybridizing the membrane with radiolabeled oligonucleotide probes. However, mature microRNA molecules are very short and their prevalence in total RNA is also very low, leading to the poor sensitivity of routine northern analysis. Several labs have developed novel northern blot methods with improved sensitivity for microRNA detection [32, 33]. Valoczi et al. applied LNA (locked nucleic acid)-modified oligonucleotide probes to northern analysis [32]. LNA is a special nucleotide whose ribose backbone is chemically modified, resulting in a greater affinity between LNA probes and target RNA. It was found that the sensitivity of this type of probe was at least tenfold higher than that of standard DNA probes [32]. The method of Pall et al. increased the efficiency with which the microRNA was fixed onto the membrane by using soluble carbodiimide to crosslink RNA, which provided a 25–50 fold increase in microRNA detection sensitivity compared to the traditional UV crosslinking method [33]. Another method widely used to detect microRNA expression is real-time RT-PCR, which can quantify specific microRNAs in samples [34, 35]. However, the similar sizes of mature microRNAs and standard PCR primers limits the direct application of conventional RT-PCR protocols to miRNA detection. To solve this problem, Applied Biosystems Co. applied a stem loop primer to a real-time RT-PCR system [34]. Their scheme included two main parts: first, the stem-loop primer was hybridized with microRNA and then reverse transcription was employed; second, a PCR primer pair was designed according to the mature microRNA sequence itself (forward primer) and the unfolded stem-loop primer (reverse primer), and then real time PCR was performed with a TaqMan probe. The advantages of this primer are as follows: 1) the base stacking of the stem can enhance the thermal stability of the primer/RNA duplex, which improves on the RT efficiency of the relatively short RT primers; 2) the double-stranded structure of the stem-loop primer prevents its nonspecific binding to other RNA molecules and thus enhances the specificity of this method. Owing to the high sensitivity and specificity of this method, microRNAs in a single stem cell could even be detected [34]. MicroRNA ISH is an important detection method that can provide information on the location of miRNA expressed in cells or tissue as well as the miRNA abundance [36, 37]. However, the normal DNA or RNA probes may not work well in this method owing to their poor binding affinity to target microRNA. To increase the affinity, Wienholds et al. introduced LNA into the ISH probes and successfully observed the expression of 115 conserved vertebrate microRNAs in zebrafish embryos by ISH [36]. They found that most microRNAs were expressed in a tissue-specific manner during segmentation and the later stages, but not early in development, which demonstrated that their function was not involved in tissue fate establishment but in differentiation or maintenance of tissue identity [36].

The methodologies mentioned above have their own advantages and disadvantages. For instance, northern blotting is considered the “gold standard” of microRNA detection, but it is very time consuming and requires large amounts of RNA samples and radioactive probes; cloning has the advantage of identifying new microRNAs, but it is not an accurate approach for microRNA quantification. RT-PCR exhibits high sensitivity due to PCR amplification, but it is limited by high cost. ISH can precisely locate a specific miRNA within tissue, but it is not suitable for high-throughput profiling. Besides the limitations mentioned above, all of these methods have a common weakness of low throughput and slow analysis speed. Fortunately, microRNA microarrays can overcome these drawbacks by offering rapid, parallel, and high-throughput analysis. Many different microRNA microarray platforms have already been successfully applied for microRNA studies [3849].

MicroRNA microarrays

MicroRNA microarray technology is actually based on nucleic acid hybridization between target molecules and their corresponding complementary probes. A schematic flow chart of the microRNA profiling microarray is shown in Fig. 1. MicroRNA oligonucleotide probes that usually have amine-modified 5′ termini are immobilized onto glass slides through covalent crosslinking between the amino groups and the SAM (self-assembling monolayer), forming a ready-to-use microRNA microarray. The isolated microRNAs are labeled with fluorescent dye and then hybridized with the microRNA microarray, resulting in specific binding of the labeled microRNAs to the corresponding probes. The fluorescence emission from labeled microRNAs bound at different positions on the slides can be detected. Consequently, the kinds of microRNAs and their relative quantities in the studied sample can be evaluated by analyzing the fluorescence signal data. The design of the microRNA probes, the preparation of microRNA samples and the labeling of microRNAs are considered the most important procedures in the microRNA microarray platform. We will discussed them in detail in the following sections.
https://static-content.springer.com/image/art%3A10.1007%2Fs00216-008-2570-2/MediaObjects/216_2008_2570_Fig1_HTML.gif
Fig. 1

The principle of the microRNA profiling microarray. (a) Amine-reactive glass slides. (b) Amine-modified microRNA probes that consist of “linker” sequences (in purple) and capture sequences (in green). (c) Ready-to-use slides of the microRNA microarray. (d) Samples (cells, tissues, and so on). (e) Isolated microRNAs. (f) MicroRNAs labeled with fluorescence dye

MicroRNA probe design

The oligonucleotide probes used in a microRNA microarray usually consist of two parts: “linker” sequences and capture sequences. The “linker” sequences often consist of poly(dT) or poly(dA) with an amine-modified terminus. In immobilized probes, the linker parts are close to the glass surface (see Fig. 1) to minimize spatial obstacles during the hybridization of the probes with the target molecules. The capture sequences are usually complementary in sequence to the microRNAs. In general, it is not complicated to design one or few probes to detect the corresponding microRNA according to the description above. However, because there are often hundreds to thousands of the probes in the microRNA microarray, and the hybridization is usually carried out at one temperature, the wider Tm distribution of the probes must be considered. For instance, the maximum difference in Tm for 1112 probes designed as described above for human, rat and mouse can be as large as about 30 °C (see Fig. 2). If the hybridization temperature is T, probes with Tm < T will give a lower binding efficiency, while the probes with Tm > T will have a higher efficiency, resulting in a serious distortion in the fluorescent signals. Therefore, Tm normalization of the full set of probes is absolutely required. In the common gene microarray, Tm normalization can be easily achieved by choosing the region of the gene probe targeted or adjusting the length of the probe. However, these methods are not suitable for Tm normalization of microRNA probes due to the size limitations of microRNAs. Several groups have developed different strategies to solve this problem. Castoldi et al. utilized LNA in probes to increase and normalize Tm [38]. It is known that LNA can enhance the thermal stability of microRNA/probe heteroduplexes, resulting in an increase of Tm. Therefore, the probe Tm can be normalized and increased by adjusting the LNA contents and lengths of the probes. The Tm values of most probes used for human and mouse microRNAs can be increased to as high as about 72 °C using this method. Castoldi et al. reported that their microarray, which used LNA probes, had high sensitivity and specificity [38]. Meanwhile, Baskerville et al. normalized Tm values by simply adjusting the lengths of the probes [39]. In this method, the known adaptor sequences are ligated to either one or both ends of the microRNAs during the labeling or amplification procedure. Then the probe is suitably lengthened based on the known adaptor sequence if the original Tm is too low or appropriately truncated if the original Tm is too high. Our laboratory employed this method to normalize the Tm values of 1112 probes for human, rat and mouse microRNAs in the Sanger database (version 10.0) [50]. Figure 2 shows the distribution of probe Tm after normalization. It shows that the Tm distribution (black curve) is much narrower than that obtained without normalization (green curve). All of the normalized Tm values are in the range 55 ± 3 °C.
https://static-content.springer.com/image/art%3A10.1007%2Fs00216-008-2570-2/MediaObjects/216_2008_2570_Fig2_HTML.gif
Fig. 2

Tm (melting temperature) distribution for microRNA probes for human, rat and mouse. Red and black curves represent the Tm distributions of the raw and normalized probes, respectively

Preparation of microRNA

The quality of RNA is important in microRNA microarray experiments. Most commercialized reagents used for total RNA extraction can meet the requirements of microRNA microarray experiments. As the abundance of microRNAs in total RNA is very low, it is best to enrich and isolate microRNAs from total RNA. Many commercialized kits are available for the enrichment of small RNA. The mirVana™ microRNA isolation kit of Ambion, Inc. (St. Austin, TX, USA) and the PureLink™ microRNA isolation kit of Invitrogen Co. (Carlsbad, CA, USA), which are based on the selective binding of small RNA molecules to a silica-based filter in the presence of a high concentration of ethanol, can be used to rapidly enrich low molecular weight RNAs that are less than 200 nt in length. To eliminate interference from microRNA precursors, denaturing polyacrylamide gel electrophoresis or the FlashPAGE™ Fractionator produced by Ambion, Inc. are often used to further isolate the mature microRNA fraction.

MicroRNA labeling

Unlike mRNA, microRNA cannot be labeled through poly(T)-driven reverse transcription, as microRNA molecules do not contain poly(A) tails. Thus, the effective labeling of microRNA is a key part of a microRNA microarray experiment. Many different labeling methods have been developed, and they can be classified into two main categories: 1) direct labeling, in which microRNA molecules are directly conjugated with fluorescent dye; 2) indirect labeling, in which the microRNA reverse transcript, the RT-PCR product of microRNA or the in vitro transcript of microRNA are labeled instead of the microRNA molecule itself.

Direct labeling methods

Direct labeling methods are easy to perform, and the artificial error introduced by reverse transcription and PCR amplification can be avoided. Thus, this type of labeling method has been widely applied in many labs.

Labeling guanine of microRNA

A guanine labeling reagent such as Ulysis Alexa Fluor 546/647 (Molecular Probes, Eugene, OR, USA) can be used to label microRNA, as the majority of microRNAs have guanosines. The advantage of this method is that each microRNA molecule can be labeled with more fluorescent dye because microRNA often contains several guanosines, resulting in a stronger fluorescence signal. The Hughes lab used Ulysis Alexa Fluor to label 7 μg total RNA, and obtained a result that agreed well with the northern blotting result reported by Sempere [31, 40]. Because the fluorescence signal is related to the guanosine content of the microRNA and this content varies depending on the microRNA, the signals obtained cannot be simply used to compare different microRNAs in terms of the level of expression, which could be a disadvantage of this method.

Labeling through T4 RNA ligase

This method uses T4 RNA ligase to add a fluorescence-modified (di)nucleotide (for example pCU-Cy3) onto the 3′ end of microRNA. The Hammond lab used this enzyme to ligate pCU-Cy3 to RNAs of different sizes (21–30 nt, 30–63 nt, 63–76 nt, >76 nt), including microRNA [41]. Interestingly, they found that only 18–30 bp RNAs were usually labeled and gave an appreciable array signal in detection. This phenomenon was attributed to the need for characteristic 3′-OH termini of mature microRNAs in the ligation reaction [41]. It appears that the microRNA isolation procedure can be avoided when using this labeling method.

Labeling microRNA through PAP

In this method, a poly(U) tail with amine-modified UTP is first appended to the 3′ end of microRNAs using the poly(A) polymerase(PAP) enzyme. The tailed microRNAs are subsequently labeled with amine-reactive fluorescent dyes [42].The major advantage of this labeling method is that more amino-modified nucleotides can been incorporated into each microRNA molecule, which enhances the detection sensitivity greatly. Furthermore, the number of fluorescent dye labels added to all microRNA molecules is the same, so the fluorescence signal can be used to conveniently compare microRNA abundance.

Labeling microRNA through chemical methods

This type of method directly labels the adjacent 3′-OH of mature microRNAs using a chemical reagent [43, 44]. In Ruan’s lab, microRNAs were oxidized with sodium periodate to convert the 3′ terminal adjacent hydroxyl groups into dialdehyde, which was then reacted with biotin-X-hydrazide through a condensation reaction, resulting in biotinylated microRNAs [43]. Then a novel fluorescent molecule—a quantum dot—was specifically bound to the biotinylated microRNA through streptavidin (Fig. 3). The detection limit of this microRNA microarray platform was 0.1 fmol, and the dynamic range was about 2–3 orders of magnitude. The major advantage of this labeling method is that the direct labeling of the 3′ termini of microRNA can avoid base bias in the labeling, and thus different microRNAs have similar labeling efficiencies. In addition, the application of quantum dots—which have high extinction coefficients and high fluorescent quantum yields—could enhance detection sensitivity. Based the same chemical labeling method, a rapid detection method for microRNA based on the use of colorimetric gold–silver enhancement was also developed in Ruan’s lab, which made it straightforward to detect the microarray signal with an ordinary CCD digital camera mounted on an optical microscope. This method avoids the need for costly laser scanners and thereby offers a low-cost detection alternative.
https://static-content.springer.com/image/art%3A10.1007%2Fs00216-008-2570-2/MediaObjects/216_2008_2570_Fig3_HTML.gif
Fig. 3A–B

Schematic principles of the microRNA profiling microarray based on using chemical labeling and quantum dots [43]. A Principle of direct labeling of microRNA at the 3′ terminus with biotin. B Principle of the microRNA profiling microarray detected with quantum dots or a colorimetric method

Labeling microRNA through RNA-primed array-based Klenow enzyme assay (RAKE)

In this method, the miRNA probes are different from those mentioned above, in that three thymidine nucleotides are inserted between the linker and capture parts [45]. After hybridization, the probe and its corresponding miRNA form a heterogeneous duplex which cannot be truncated by exonuclease I. Meanwhile, the probes that are not bound by miRNA molecules will be subsequently truncated by exonuclease I. In the subsequent Klenow-driven extension reaction, the microRNAs and probes in the duplex are primer and template, respectively. The biotin-labeled dATP can be incorporated into the extended part of the miRNA. Subsequently, the biotinylated miRNA can be labeled by streptavidin conjugated with fluorescent dye. The detection limit of this approach can be as high as 0.2 fmol microRNA, and the dynamic range can reach three orders of magnitude [45], which is comparable to northern blotting. Due to the need for an accurate match between the 3′ ends of the microRNAs and the corresponding probes in RAKE, this method has the special advantage that it can distinguish between paralogous microRNAs that differ at the 3′ end.

Indirect labeling methods

Considering the low abundance and stability of miRNA, indirect labeling methods have also been developed. In this type of labeling, the target for labeling is not the miRNA molecule itself, but its reverse transcript or RT-PCR products. These methods are actually well established and are widely used in molecular biology. The advantages of indirect labeling methods are: 1) the cDNAs of microRNAs are more stable than miRNA and are easy to preserve; 2) the target molecules derived from microRNAs can be amplified and labeled synchronously through PCR or in vitro transcription, which is very useful for the detection of low-abundance microRNAs.

Labeling of the microRNA reverse transcript

This method labels the reverse transcripts of the microRNAs, which are acquired through random primer-driven reverse transcription [21, 46, 47]. Liu et al. applied an eight-nucleotide random primer labeled with two biotin molecules (3′-(N)8-(A)12-biotin-(A)12-biotin-5′) to drive the reverse transcription of the total RNA template in order to obtain cDNAs labeled with biotin [21, 46]. Meanwhile, an unlabeled seven-nucleotide random primer was used in the method of Sun et al., and the cDNA obtained was then labeled at the 3′ end with the biotin-dideoxy-UTP through the catalysis of terminal transferase [47]. Due to the low content of microRNA in total RNA and the poor specificity of the random primer, this method may be prone to introducing artificial errors. It should also be noted that the design of the microRNA probes used in this method is different from that used in direct labeling methods; the capture sequences correspond to the microRNAs instead of sequences complementary to the microDNAs.

Labeling of RT-PCR products of microRNAs

In these labeling methods, two adaptor sequences are first added to the 3′ and 5′ ends of microRNA, respectively, and a biotin- or fluorescence-modified primer is designed (according to the known adaptor sequence) in order to achieve RT-PCR amplification, through which the labeled PCR products can be obtained. The advantage of this approach is that the copy number of the cDNA of the microRNA can be amplified and the labeling is carried out automatically in PCR, which can increase the detection sensitivity. The Horvitz lab employed a Cy3-labeled primer to label the sense strand of PCR product, and then the denatured PCR products were hybridized with the microarray [48]. One weakness of this method is that the presence of antisense strands in the hybridization procedure may interfere with the hybridization between sense strands and the corresponding probes. To minimize this interference, the Bartel lab applied two adapters with different lengths, where the longer one for the sense strand was fluorescently labeled, resulting in the introduction of a length discrepancy between sense and antisense strands [39]. The fluorescently labeled sense strands can be then easily isolated by denaturing polyacrylamide gel electrophoresis, so that they can be hybridized with microarrays without interference from the antisense stands. In addition, Bartel et al. use a set of DNA oligonucleotides that are complementary in sequence to all of the probes as references. The reference molecules and the RT-PCR products of microRNAs were labeled with cy5 and cy3, respectively, and they were mixed together and hybridized with the microarrays so that the reference set provided an uniform positive control and an internal standard for normalization [39].

Labeling in vitro transcript of microRNAs

In this method, PCR products were obtained as described above. However, a promoter sequence was introduced into the adaptor sequence. The in vitro transcript can then be labeled with a fluorescence-modified nucleotide (e.g., Cy3-CTP) in the transcription procedure. The advantage of this approach is that the target molecules are amplified linearly, which can enhance the copy number of the target molecules and avoid PCR amplification, which is prone to introducing artificial errors. The Rosetta lab developed this method to label microRNAs [49]. They introduced a T7 promotor sequence into the adaptor sequence, and subsequently performed further in vitro transcription through the T7 promotor. The cRNAs of the microRNAs were labeled with Cy3-CTPs or Cy5-CTPs, which were incorporated into cRNA during the in vitro transcription procedure. They found an overall agreement between the results from the array and those given in previous reports [27, 31, 51]. One point to note is that the fluorescence signal is related to the number of C residues in the microRNA; the signals obtained cannot be simply used to compare the expression levels of different microRNAs.

Conclusion

The main conventional microRNA detection methodologies, especially microRNA microarray methods, have been briefly summarized above. Although each of these methods has their own unique advantages, they have not been perfected yet. For instance, in different labeling methods for miRNA, the labeling of RT-PCR products can introduce artificial errors during the ligation and PCR amplification procedures; guanine labeling is not suitable for microRNAs lacking G residues, T4 ligase labeling is limited by a poor reputation for reliability and the base bias of the RNA ligase; the procedure of chemical labeling is a little complicated, and the degree of oxidation of the hydroxy group adjacent to the 3′ terminal is not easy to control, and so on. Therefore, it is imperative to improve the current methods mentioned above and to develop a new, ideal approach with a higher sensitivity and a high throughput for miRNA detection. However, at present, the method you choose for miRNA detection should best fit your experience, the experimental conditions in the laboratory, and the goal of your research. For instance, a cloning method might be a good choice when analyzing the microRNAs of an organism if little is known about its microRNA sequences, because the microRNA sequences are not required in this method. ISH might be the most suitable approach if the abundance(s) of one or a few specific microRNAs is/are being studied, as well as where they are expressed in tissue. Meanwhile, a microRNA microarray is the best choice when studying the expression levels of hundreds of microRNAs at same time. However, it should be noted that northern blotting is the standard technique for quantifying miRNAs in current methods. It should also be noted that there are actually some other microRNA detection methods besides those mentioned above, such as a microbead-based assay [49, 52], a single-molecule method based on two-color coincident detection by fluorescence correlation spectroscopy (FCS) [53], a molecular beacon technique [54], enzyme-based miRNA detection [55], and so on. These methods also very useful in some experiments, although they have not been as widely used as, for example, miRNA microarrays. Interested readers can find out more about them from the corresponding literature.

Acknowledgements

This work was funded by the National Key Basic Research and Development Program of China (No. 2007CB935702)

Copyright information

© Springer-Verlag 2009